Research On Interactive Control System Based On Image Denoising And Target Tracking

With the rapid development of computer vision technology,a variety of human-computer interaction devices have entered the field of family life and services.Efficient and natural interaction methods have brought great convenience and value to human life and development.However,existing Human-computer interaction technology has major problems in practical applications.For example,the digital image of the interactive system will generate various interferences during the generation,transmission,and processing,which will deteriorate the image quality,which will affect the subsequent image analysis and recognition.In addition,due to complex changes in the interactive environment,problems such as inaccurate information fusion,loss of targets,drift of anchor points,etc.are caused.In order to ensure the natural and high efficiency of the actual human-computer interaction process,for the above two problems,this paper will improve the interaction performance by improving the method from the image denoising and target tracking aspects.According to the research and analysis of the noise interference of interactive systems,it is found that most human-computer interactive systems are generally subject to two types of noise interference,Gaussian noise and salt and pepper noise.An improved noise reduction algorithm that is effective at the same time is proposed.In addition,this paper addresses the problems of inaccurate information fusion,loss of moving targets,and information holes caused by the inability to obtain the target template in the interactive system.This paper also proposes an improved algorithm for target tracking based on fused information.The specific content results and innovations are as follows:(1)In order to solve the problems of image quality deterioration and insufficient information fusion caused by interference noise,this paper proposes to use an improved noise reduction algorithm based on non-local mean algorithm in the image preprocessing stage.Non-Local Means has the advantages of low time complexity and good edge information retention during denoising,but there is also a single limitation to the suppression of noise types.In this paper,based on the analysis of the principle of the denoising algorithm,the improvement strategy of the NLM algorithm for the above two noises is proposed above.The experiment proves that the algorithm can be effective for Gaussian noise and salt and pepper noise in the interactive system at the same time,and the pre-processed image interference Significantly reduced,the image is clearer.(2)Aiming at the problems of target loss,information hole and interactive drift caused by the target being blocked or the environment changing in the interactive system,this paper proposes a motion tracking algorithm combining Kalman filter and Camshift based on the fusion of image information in the moving target tracking stage,And put forward an improved strategy of position prediction to measure the degree of occlusion.The algorithm linearly fits the output results of Kalman and Camshift,which has the advantages of high real-time performance,ability to estimate motion position,and fast calculation speed.Finally,multiple experiments are used to prove that the tracking algorithm proposed in this paper is faced with occlusion and other environmental changes Can still maintain good robustness,which can effectively improve the tracking performance of the interactive system.(3)A set of non-contact interactive system is designed and built using lidar and visual sensors.Fusion of lidar depth information and visual image information was analyzed and improved from the image denoising algorithm and target tracking algorithm respectively.Finally,the key technologies such as image denoising,information fusion and target tracking were combined through the three-dimensional interactive software Ventuz.to optimize the performance of interactive systems.